Audio APIs

Live
Synthesize speech or transcribe uploaded and hosted audio.
Open in ChatGPT
(opens in new tab)Last verified: 2026-09-08

Text to speech

Send up to 600 characters. The endpoint defaults to WAV; use response_format for another format supported by the model. Check GET /models for voice, speed and format limits.
bash
curl https://api.modelrush.ai/v1/audio/speech \
  -H "Authorization: Bearer $MODELRUSH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"modelrush/qwen3-tts-flash","input":"ModelRush connected","voice":"Cherry"}' \
  --output speech.wav

Speech to text

Submit multipart audio or JSON containing a supported public audio reference. The response includes transcribed text, model, region, and request identifiers.
bash
curl https://api.modelrush.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer $MODELRUSH_API_KEY" \
  -F "model=modelrush/qwen3-asr-flash" \
  -F "file=@meeting.wav"

Voice enrollment

For Qwen Voice Enrollment create, optional language describes the reference audio. Choose the language actually spoken; if omitted, the provider defaults to Chinese. Supported codes: zh / en / de / it / pt / es / ja / ko / fr / ru. Use POST /v1/voices with model: "modelrush/qwen-voice-enrollment", action: "create", audio, target_model and preferred_name. Omit language to keep the provider default. An unsupported code returns HTTP 400 with unsupported_voice_language before a billable request is created.
PreviousVideo generation
NextRegions and pricing